fMRI of visual encoding: Reproducibility of activation

Abstract

fMRI, a noninvasive technique to measure brain activation, is gaining clinical interest, because its sensitivity enables individual assessments. However, more insight in the reproducibility of these measurements during higher cognitive tasks is necessary. We performed an fMRI study involving withinand between-subject reproducibility during encoding of complex visual pictures. Ten healthy subjects were studied on three occasions: twice in the same scanning session (study 1 and 2), and a third time, 3-24 days later (study 3). On all 30 occasions but one, activation was found in areas expected on the basis of previous studies, including the fusiform and lingual gyri, occipital and parietal areas, the (para)hippocampal area, and the frontal inferior sulcus. The reproducibility of the number of activated voxels in the whole brain was 72% and 63% (respectively, studies 1 and 2, and 1 and 3). The reproducibility of anatomical identical pixels that supplement these results was 49% and 36%. These reproducibility measures increase about 5-15% when only areas of expected activation are included. The quantitative measurements indicate that there is substantial variation in the volume of activation. The recognition of pictures as tested afterward explains part of this variation between subjects. Our findings indicate that whereas consistent patterns of activation exist, more insight is needed into what determines the volume of activation, especially to assess cognitive alterations in patients over time. Hum. Brain Mapping 9:156 -164, 2000.

A re-defined and generalized percent-overlap-of-activation measure for studies of fMRI reproducibility and its use in identifying outlier activation maps

“…This is also true of the pairwise percent overlap measure of activation proposed by Rombouts et al (1998) and Machielsen et al (2000).…”

“…This paper, therefore, introduces and studies its use in quantifying fMRI reproducibility in Section 2.1. This measure, although a slight modification to the Rombouts et al (1998) and Machielsen et al (2000) definition of ω j,l , is seen to be both intuitive and physically interpretable. At the same time, like ω j,l , it is also a pairwise reliability measure so that we get M 2 overlap measures ω j,l , 1 ≤ l ≤ j ≤ M from M fMRI studies.…”

“…These maps can vary greatly from one study to the next, so quantifying the reliability of identified activation over several fMRI studies is important. The percent overlap of activation (Rombouts et al, 1998 andMachielsen et al, 2000) is a global reliability measure between activation maps drawn from any two fMRI studies. A slightly modified but more intuitive measure is provided by the Jaccard (1901) coefficient of similarity, whose use we study in this paper.…”

“…In these scenarios, Rombouts et al (1998) and Machielsen et al (2000) have proposed a global reliability measure for any pair of fMRI studies: For any two replications (say, j and l), the percent overlap of activation is defined as ω j,l = 2V j,l /(V j +V l ), where V j,l is the number of three-dimensional image voxels identified as activated in both the jth and the lth replications, and V j and V l represent the number of voxels identified as activated in the jth and the lth experiments, respectively. Thus, ω j,l is a ratio of the number of voxels identified as activated in both replications to the average number of voxels identified as activated in each replication.…”

Evidencing a place for the hippocampus within the core scene processing network

“…Here, we address this question by conducting a large‐scale analysis of functional localiser data with the aim of providing information about individual‐level activations [i.e., the number of participants with scene‐sensitive voxels in HC relative to other brain areas; see Machielsen and Rombouts, 2000], but also understanding the spatial profile of these individual‐level activations and how it compares to group‐level statistics [e.g., Engell and McCarthy, 2013; Morrison and Downing, 2007; Nieto‐Castañón and Fedorenko, 2012]. While several studies have attempted to characterise the individual‐level consistency of scene‐sensitive brain activations, these studies have either used small sample sizes [Nasr et al, 2011; Spiridon et al, 2006], or focussed on a single scene processing region [e.g., Peelen and Downing, 2005].…”